--- Martin Lichtin <lichtin@xxxxxxxxx> wrote:
> > I'm having the following problem with RH7.1
> Kickstarts:
> >
> > During the partitioning phase, when mke2fs runs
> for a reasonably large
> > partition (eg. 10-100GB), mke2efs often, but not
> always, hangs at the
> > following point:
> >
> > # mke2fs /dev/rd/c0d0p2
> > [mke2fs output...]
> >
> > Writing inode tables: 1/683
> >
> > It just sits there, sometimes forever (I once
> cancelled kickstart
> > after a day).
>
> Well, it turns out that if I create the partition
> "manually" in the
> %post section, it works just fine. There's something
> weird about the
> way anaconda runs mke2fs, it seems.
>
We're seeing a similar problem. Our situation:
mke2fs running on an LVM-ized partition running on
top of an IBM serverraid partition. (Why, you might
ask, bother running LVM on top of hardware RAID?
Because some of our systems won't have RAID,
and adding more disk space to a filesystem
means adding another disk, and you need LVM to make
that work. And we don't want to have one release for
non-RAID systems, and another release for RAIDed
systems, if we can avoid it.)
We're seeing it hang somewhere above 36 Gbyte. We
have
not had a chance to characterize it further, or to see
if the problem goes away with Martin's solution of
doing it during %post.
What are some of the differences between ramdisk
and %post?
1. df works in ramdisk. In post it says it can't
find a table its looking for.
2. There are no /dev entries in ramdisk. It's
possible to use mknod to construct them. One
of the LVM tools uses the /dev entries (e.g.,
/dev/sda, /dev/sda1, /dev/sda2, and so on.)
3. All the usual binaries are in /bin, /usr/bin, and
so on in %post.
Martin, when you run the mke2fs "manually" in %post,
is it a %post --nochroot or a regular %post? If it's
a regular %post, what happens if you try it with a
%post --chroot?
--Seth
sethal@xxxxxxxxx
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail
http://personal.mail.yahoo.com/
